The B2B startup Restb.ai is located in Barcelona and its philosophy is to customize computer vision. The team is promoting machine learning algorithms for companies that want to fix weak points or want to improve customer service quality by automatically identifying specific images.
Although tech giants like Facebook and Google are using their own data to train their artificial intelligence algorithms in order to achieve their goals. But Restb.ai's vision is to sell computer vision as a service, first charging its business customers for initial training fees, then building the custom models they need, and finally charging the APIs one by one.
VP Damia Garcia of the growth and business development team said: The team was established in March 2015 and currently has six customers. He added that its early acquisition efforts focused on acquiring large customers in multiple vertical markets, and now the goal is to expand the audience.
Currently, one of the users of this technology is the second-hand goods market Wallapop, which has used Restb.ai to identify and prevent users from posting illegal drug images on the site for sale.
Garcia also emphasized the accuracy of the Restb.ai model, which is 98% accurate once the template has been trained, regardless of the customer's intended customization.
The working principle of this system is: the user inputs any custom keywords that have been trained on the model, such as the automobile manufacturing/model in the above example, and then the model will give the similarity to the specified identifier, the working process Close to real time.
Garcia also pointed out another application of the technology. The police used automakers/models to automatically cross-reference the license plate to ensure that the license plate was not changed to another car.
He said: "The training process for each custom model takes about three weeks, and because of the different application models, artificial intelligence systems need to input large amounts of data. For 150-2000 images, you usually need to train an incidental customer to provide The data, specific custom requirements."
Although the algorithm only requires a photo of a taillight (for example), or a photo of a car accident, it can identify the model of the car. However, in the demo car example, each car requires 1500 pictures. It can also specify that the vehicle can be determined from that side/perspective.
Other examples of using Restb.ai computer vision as a service are:
• The real estate market wants to identify and index photos within an apartment, for example, processing images that need to be manually tagged for uploads, and provide better content goals based on better classification of image data.
· Norwegian mining company, also makes use of high-tech automation to quickly identify specific mineral crystals, some of the staff assigned to carry out other tasks.
Garcia found a competitor in the United States—Clarifai—but their main target was general modeling, such as identifying nudity, and Restb.ai wanted to provide custom deployments that met specific customer differentiation needs.
Via:TC
Recommended reading:
Intel Acquires Movidius to Advance in Computer Vision
One billion is a lot? Domestic AI companies have completed this small goal
Trench Cover,Trench Drain,Drain Cover,Steel Trench Covers
Hunan Furui Mechanical and Electrical Equipment Manufacturing Co., Ltd. , https://www.thresher.nl